diff --git a/cudart64_110.dll b/cudart64_110.dll new file mode 100644 index 000000000..4e7c59c16 Binary files /dev/null and b/cudart64_110.dll differ diff --git a/make_old_pyinstaller_cuda.bat b/make_old_pyinstaller_cuda.bat index 28a03c771..a02a77ac7 100644 --- a/make_old_pyinstaller_cuda.bat +++ b/make_old_pyinstaller_cuda.bat @@ -1,4 +1,4 @@ echo This file is only for my own usage, please do not use it. I am lazy. set PATH=d:\\MainApplications\\KoboldAIGPT\\KoboldAI-Horde-Bridge\\python;d:\\MainApplications\\KoboldAIGPT\\KoboldAI-Horde-Bridge\\python\\Scripts;%PATH% -PyInstaller --noconfirm --onefile --clean --console --icon "./niko.ico" --add-data "./klite.embd;." --add-data "./koboldcpp.dll;." --add-data "./cublas64_11.dll;." --add-data "./cublasLt64_11.dll;." --add-data "./cudart64_110.dll;." --add-data "./msvcp140.dll;." --add-data "./vcruntime140.dll;." --add-data "./vcruntime140_1.dll;." --add-data "./rwkv_vocab.embd;." --add-data "./rwkv_world_vocab.embd;." "./koboldcpp.py" -n "koboldcpp.exe" \ No newline at end of file +PyInstaller --noconfirm --onefile --clean --console --icon "./nikogreen.ico" --add-data "./klite.embd;." --add-data "./koboldcpp.dll;." --add-data "./koboldcpp_openblas.dll;." --add-data "./koboldcpp_failsafe.dll;." --add-data "./koboldcpp_openblas_noavx2.dll;." --add-data "./libopenblas.dll;." --add-data "./koboldcpp_clblast.dll;." --add-data "./clblast.dll;." --add-data "./koboldcpp_cublas.dll;." --add-data "./cublas64_11.dll;." --add-data "./cublasLt64_11.dll;." --add-data "./cudart64_110.dll;." --add-data "./msvcp140.dll;." --add-data "./vcruntime140.dll;." --add-data "./vcruntime140_1.dll;." --add-data "./rwkv_vocab.embd;." --add-data "./rwkv_world_vocab.embd;." "./koboldcpp.py" -n "koboldcpp.exe" \ No newline at end of file diff --git a/msvcp140.dll b/msvcp140.dll new file mode 100644 index 000000000..5b83ea2b7 Binary files /dev/null and b/msvcp140.dll differ diff --git a/nikogreen.ico b/nikogreen.ico new file mode 100644 index 000000000..636222311 Binary files /dev/null and b/nikogreen.ico differ diff --git a/otherarch/llama_v2.cpp b/otherarch/llama_v2.cpp index 2f8e168ca..1dee94be7 100644 --- a/otherarch/llama_v2.cpp +++ b/otherarch/llama_v2.cpp @@ -1063,6 +1063,8 @@ static void llama_v2_model_load_internal( #if defined(GGML_USE_CUBLAS) { const int n_gpu = std::min(n_gpu_layers, int(hparams.n_layer)); + if(GetQuantsUnshuffled()) + { fprintf(stderr, "%s: [old cublas] offloading %d layers to GPU\n", __func__, n_gpu); @@ -1085,6 +1087,14 @@ static void llama_v2_model_load_internal( } fprintf(stderr, "%s: [old cublas] total VRAM used: %zu MB\n", __func__, vram_total / 1024 / 1024); + } + else + { + if(n_gpu>0) + { + printf("\n[WARNING: Old format does not support GPU offloading! It will be deactivated!]\n"); + } + } } #elif defined(GGML_USE_CLBLAST) { diff --git a/vcruntime140.dll b/vcruntime140.dll new file mode 100644 index 000000000..a5bdfeb62 Binary files /dev/null and b/vcruntime140.dll differ diff --git a/vcruntime140_1.dll b/vcruntime140_1.dll new file mode 100644 index 000000000..6d9bbe8d8 Binary files /dev/null and b/vcruntime140_1.dll differ